Engineered disulfide bonds improve thermostability and activity of L‐isoleucine hydroxylase for efficient 4‐HIL production in Bacillus subtilis 168

Abstract 4‐Hydroxyisoleucine, a promising drug, has mainly been applied in the clinical treatment of type 2 diabetes in the pharmaceutical industry. l‐Isoleucine hydroxylase specifically converts l‐Ile to 4‐hydroxyisoleucine. However, due to its poor thermostability, the industrial production of 4‐hydroxyisoleucine has been largely restricted. In the present study, the disulfide bond in l‐isoleucine hydroxylase protein was rationally designed to improve its thermostability to facilitate industrial application. The half‐life of variant T181C was 4.03 h at 50°C, 10.27‐fold the half‐life of wild type (0.39 h). The specific enzyme activity of mutant T181C was 2.42 ± 0.08 U/mg, which was 3.56‐fold the specific enzyme activity of wild type 0.68 ± 0.06 U/mg. In addition, molecular dynamics simulation was performed to determine the reason for the improvement of thermostability. Based on five repeated batches of whole‐cell biotransformation, Bacillus subtilis 168/pMA5‐ido T181C recombinant strain produced a cumulative yield of 856.91 mM (126.11 g/L) 4‐hydroxyisoleucine, which is the highest level of productivity reported based on a microbial process. The results could facilitate industrial scale production of 4‐hydroxyisoleucine. Rational design of disulfide bond improved l‐isoleucine hydroxylase thermostability and may be suitable for protein engineering of other hydroxylases.


INTRODUCTION
4-Hydroxyisoleucine (4-HIL) is a natural nonproteinogenic amino acid that was first discovered in fenugreek seeds [1,2]. 4-HIL is a promising drug with potential applications mainly in the pharmaceutical industry. 4-HIL has attracted a great have been mainly applied in the extraction of 4-HIL from fenugreek seeds, but the yields are relatively low (approximately 150 mg of 4-HIL extracted from 1 kg of fenugreek seeds) [8]. Chemical-enzymatic synthesis methods have also been evaluated and are associated with low efficiency, high costs, and severe pollution [5][6][7]. With the constant increase in environmental awareness and the development of green technologies, nonpolluting and nontoxic biological technologies have inevitably become favored in industrial development [9,10].
Recently, the use of whole-cell processing L-isoleucine hydroxylase (IDO) to specifically catalyze L-Ile into 4-HIL has received considerable attention. In addition, ido derived from Bacillus thuringiensis 2e2, Bacillus thuringiensis TCCC 11826, Bacillus thuringiensis YBT 1520, and Bacillus weihenstephanensis has been expressed in Escherichia coli or Corynebacterium glutamicum. 4-HIL can be obtained using biotransformation or direct fermentation [1,2,[11][12][13]. However, due to its poor thermostability, IDO enzyme activity decreased considerably when the temperature exceeded the optimal value [1,2,[14][15][16]. To date, to the best of our knowledge, no study has explored or reported on the molecular modification of IDO to improve its thermostability.
The methods that have been extensively applied to improve enzyme thermostability include homology alignment, proline design, disulfide bond design, and surface charge optimization [17]. The introduction of disulfide bonds into cellobiohydrolase [18,19], lipase [20,21], and xylanase [22,23] considerably improved their thermostability and activity. The introduced disulfide bond stabilizes the protein in a reversibly folded state by decreasing the folding entropy [24] or in an irreversibly folded state by decreasing the folding rate of the protein [25]. So, we designed disulfide bonds on the loops to enhance IDO thermostability to facilitate efficient 4-HIL production.
Bacillus subtilis, an industrial workhorse that is widely applied in the large-scale microbial production of recombinant proteins, amino acids, and fine chemicals [26,27], is generally recognized as safe by the Food and Drug Administration (FDA) [28][29][30]. Although B. subtilis has been extensively applied in the production of valuable metabolites, bioremediation, and energy generation, there was no study with regard to its application in 4-HIL production. Therefore, we selected B. subtilis 168 as the host that would express IDO for 4-HIL production.
In the present study, we first expressed ido from Bacillus cereus 13658 in B. subtilis 168. The thermostability and enzyme activity of IDO were increased through the rational design of the disulfide bond. The half-life of IDO at 50 • C increased from 0.39 to 4.03 h, which was a 10.27-fold increase. In addition, the specific enzyme activity of IDO increased from 0.68 ± 0.06 U/mg to 2.42 ± 0.08 U/mg, which was a 3.56-fold increase. In addition, we investigated the

PRACTICAL APPLICATION
4-HIL has the effects of promoting insulin secretion, improving the resistance of peripheral tissues to insulin, and regulating dyslipidemia, which is mainly used in the pharmaceutical industry. L-isoleucine hydroxylase is commonly used to produce 4-HIL. However, the applicability of L-isoleucine hydroxylases in industrial applications is restricted by their low thermostability or low activity at high temperature. Therefore, we designed the disulfide bond for L-isoleucine hydroxylase to improve thermostability. Finally, the obtained optimal mutant strain Bacillus subtilis 168/pMA5-ido T181C rain produced a cumulative yield of 856.91 mM (126.11 g/L) 4-HIL after five batches, which is the highest level of productivity reported based on a microbial process. This strategy enhanced thermostability and enzyme activity of L-isoleucine hydroxylase, which potentially be used for large-scale 4-HIL industrial production. The approach also may be suitable for protein engineering of other hydroxylases for the efficient production of other valuable hydroxylated amino acids. mechanism of half-life and specific enzyme activity increase using molecular dynamics simulation.
Finally, the whole cell of B. subtilis 168/pMA5-ido T181C recombinant strain was used for the biotransformation of L-Ile into 4-HIL. The five repeated batches gave a cumulative yield of 856.91 mM (126.11 g/L) 4-HIL, which was 2.19 times higher than the 390.52 mM (57.47 g/L) yield obtained from the wild-type strain. The results of the present study offer novel insights that could facilitate the enhancement of IDO thermostability and pave the way for the industrial application of 4-HIL.

Strains, plasmids, and materials
ido (GenBank: AE016877.1) encoding IDO from B. cereus 13658 was used as the template. Plasmid pMA5 was utilized as the expression vector for IDO and its mutants. E. coli JM109 was used as the host for ido cloning and B. subtilis 168 as the expression host for IDO and its variants. PrimeS-TAR HS DNA polymerase, restriction enzymes, and T4 DNA ligase were purchased from TaKaRa (Dalian, China).

Construction of wild type strain and its mutants
The ido amplified by PCR was ligated to pMA5 plasmid to construct recombinant plasmid pMA5-ido. Site-directed mutagenesis on the K152 and T181 sites of the ido to cysteine was performed by overlap extension PCR [31] using primers listed in Supporting Information Table S1. The 6×His tag was added downstream of the primer to facilitate purification of the recombinant enzymes. The mutated genes (ido K152C and ido T181C ) were also ligated to pMA5 to generate pMA5-ido K152C and pMA5-ido T181C . Recombinant plasmids pMA5-ido, pMA5-ido K152C , and pMA5-ido T181C were transformed into E. coli JM109 to obtain recombinant strains E. coli JM109/pMA5-ido, E. coli JM109/pMA5-ido K152C , and E. coli JM109/pMA5-ido T181C , respectively. The recombinant plasmids were also transformed into B. subtilis 168 to obtain recombinant strains B. subtilis 168/pMA5-ido, B. subtilis 168/pMA5-ido K152C , and B. subtilis 168/pMA5-ido T181C (Supporting Information Table S2).

Protein expression and purification
The recombinant strains B. subtilis 168/pMA5-ido, B. subtilis 168/pMA5-ido K152C , and B. subtilis 168/pMA5-ido T181C were cultured overnight at 37 • C in Lysogeny Broth (LB) medium supplemented with 50 µg/mL kanamycin. The cells were then incubated into 50 mL LB medium and cultured at 37 • C shaker at 180 rpm for 20-24 h to obtain IDO, K151C, and T181C protein. The cells were harvested by centrifugation at 8000 rpm at 4 • C for 10 min (refrigerated centrifuge, Sigma) and then washed twice with phosphate-buffered saline (PBS) (50 mM, pH 7.0). The harvested cells were added to lysozyme for 1-2 h on ice and sonicated and centrifuged at 10 000 rpm for 30 min at 4 • C to remove cell debris. The supernatant was purified and used for the enzyme activity assay.
Wild type and its variants were subjected to metal affinity chromatography on an AKTA purifier system (Amersham Pharmacia Biotech, UK) using a 1-mL His Trap FF column (GE Life Sciences, USA) with linear gradient elution experiments using 0 to 700 mM imidazole as the elution buffer. The purified wild type and its mutant enzymes were analyzed by SDS-PAGE (12% acrylamide). Protein concentrations were determined using a Bradford Protein Assay Kit [32].

Enzyme activity assay and 4-HIL determination
Wild type and its mutants' enzyme activities were determined using the 2,4-dinitrofluorobenzene precolumn derivatization method [1]. The assay mixture contained 30 mM L-isoleucine, 30 mM -KG, 1 mM FeSO 4 ⋅7 H 2 O, and 5 mM ascorbic acid in PBS buffer (50 mM, pH 7.0) at 30 • C for 30 min with 0.1 mg/mL of wild type and its mutants. The reaction solution was boiled in a water bath for 5 min to terminate the enzyme reaction. The reaction solution was subsequently subjected to the derivatization treatment, and the 4-hydroxyisoleucine concentrations were determined using HPLC. The amount of enzyme that converted L-Ile into 1 µmol 4-HIL was defined as 1 unit (U).

Determination of optimum temperatures and pH of the wild type enzyme and its variants
The optimum temperatures for wild type and its mutants were determined by adding 0.1 mg/mL enzymes to a reaction system consisting of 30 mM L-Ile, 30 mM -KG, 1 mM FeSO 4 ⋅7H 2 O, 5 mM ascorbic acid, and PBS buffer (50 mM, pH 7.0). Relative enzyme activity was measured at 20-60 • C for 1 h. The optimum pH was determined by adding 30 mM L-Ile, 30 mM -KG, 1 mM FeSO 4 ⋅7H 2 O, 5 mM ascorbic acid, and 0.1 mg/mL of the wild type enzyme and its mutants in buffers with different pH values (3.0-10.0). The reaction was carried out at 30 • C for 1 h, and the relative enzyme activity was measured. The different buffers used included the following: 50 mM sodium citratecitrate buffer (pH 3.0-5.0), PBS buffer (pH 6.0-8.0), Tris-HCl (pH 9.0), and sodium bicarbonate-sodium carbonate solution (pH 10.0).
The thermostability of wild type and its mutants was determined by incubating them at 10-70 • C for 20 min, prior to the addition of 30 mM L-Ile, 30 mM -KG, 1 mM FeSO 4 ⋅7H 2 O, 5 mM ascorbic acid, and PBS buffer (50 mM, pH 7.0). The reaction mixture was incubated at 30 • C for 1 h, and the residual enzyme activity was measured. To further examine the thermostability of wild type and its variants, the half-life (t 1/2 ) of the wild type and its variants were tested at 50 • C. Here, similar concentrations of wild type and its variants were heat treated at 50 • C for different time periods, then added to a reaction system consisting of 30 mM L-Ile, 30 mM -KG, 1 mM FeSO 4 ⋅7H 2 O, 5 mM ascorbic acid, and PBS buffer (50 mM, pH 7.0) and reacted at 30 • C for 1 h prior to the measurement of the relative enzyme activity. The first-order rate constant k was obtained from the slope of the incubation time and the semilogarithmic plot of residual activity, and the t 1/2 value at 50 • C was calculated using the following equation: t 1/2 = ln2/k. Enzyme incubations were performed in metal bath at a constant temperature to precisely control temperature and incubation time.
The kinetic constants (K m , k cat , and k cat /K m ) were measured under conditions limiting only one substrate (L-Ile or -KG). The K m of -KG was measured based on a change in the -KG concentration from 0.0125 to 40 mM at an L-Ile concentration of 20 mM, and the L-Ile concentration was determined based on a change from 0 to 5 mM when the -KG concentration was 20 mM. Finally, the corresponding data were substituted in the Michaelis-Menten equation to obtain the kinetic constants k cat and K m of the two substrates using GraphPad Prism v7.0 (GraphPad Software, Inc., San Diego, CA, USA). All kinetic constants are average values obtained from triplicate measurements.

Verification of disulfide bond formation
To confirm the presence of a disulfide bond in the variant protein, the formation of a disulfide bond was determined by analyzing the concentrations of free cysteine in the protein using dithionitrobenzoic acid as a reagent [33].

Circular dichroism for protein secondary structure analysis
Wild type and its variants were scanned in a quartz cuvette with a diameter of 0.1 cm using a circular dichroism (CD) instrument (MOS-450/AF-CD-STP-A, Bio-Logic, Grenoble, France) in the far ultraviolet region in the range of 190-260 nm [34]. Ultrapure water was used as the control. The data from the CD scan was processed using the DichroWeb online server (http://dichroweb.cryst.bbk.ac.uk/html/process.shtml) to determine the percentage estimates of the -helix, -sheet, -turn, and random curl contents [35].

Whole-cell biotransformation of 4-HIL
The one-batch whole-cell biotransformation approach applied involved dosing of the substrate and its cofactors for 4-HIL production. Here, 200 mM L-Ile and -KG, 5 mM FeSO 4 ⋅7H 2 O, 10 mM ascorbic acid, and 200 mL PBS buffer (50 mM, pH 7.0) were added into a 500-mL shake flask, and then 10 g cells and 1 mL of Triton X-100 were added, successively. The transformation was carried out at 30 • C, and samples were obtained every 3 h over a 21-h period conversion.

Selection of the IDO target residues for mutation
The properties of the pure wild type enzyme were studied. The optimum reaction temperature was 30 • C. When the temperature was increased to 40 • C, enzyme activity decreased significantly to 52.6% (Supporting Information Figure S1A), which could limit the industrial production of 4-HIL. Therefore, it is critical to improve the thermostability of IDO using rational modification.
The amino acid sequences of IDO were submitted to Phyre2 to predict the 3D-structure model. The crystal structure of the protein (PDB: 3pl0B) was used as a template for modeling, and 226 residues (92% of the sequence) were modeled with 100% confidence using the single highest scoring template. The quality of the predicted structure was evaluated using the SAVES v5.0 [37,38]. A Ramachandran plot revealed that 97.1% of the residues were in allowed regions and 0.4% in disallowed regions. WHATCHECK, PROVE (55 buried outlier protein atoms, 7.2% error), and ERRAT (Overall Quality Factor A: 87.4) analyses also confirmed that the model was reliable. Therefore, based on all the analytical results,

Properties of wild type and its mutants
There were no significant differences in the optimum temperatures (30 • C) and pH (7.0) of wild type and its mutants (Supporting Information Figure S1). The relative enzyme activity of T181C (181 threonine residues of L-isoleucine hydroxylase protein mutated to cysteine) and K152C (152 lysine residues of L-isoleucine hydroxylase protein mutated to cysteine) remained at 84.3% and 90.3%, respectively, at 40 • C, which was an improvement compared with the 52.6% in wild type. To verify their thermostability, wild type and its mutants were incubated at 10-70 • C for 20 min to measure residual enzyme activity. T181C retained 69.4% of enzyme activity, whereas wild type and K152C nearly lost enzyme activity ( Figure 1A). To further reveal the differences in thermostability, these enzymes were incubated at 50 • C for different lengths of time, and the half-life of heat inactivation (t 1/2 ) was determined ( Figure 1B). According to the results, T181C was considerably more stable compared with both wild type and K152C. In addition, the inactivation process was consistent with the first-order kinetics (Supporting Information Figure S3), the The specific activity of T181C was 2.42 ± 0.08 U/mg, which was 3.56-fold that of wild type, which was 0.68 ± 0.06 U/mg, whereas that of K152C decreased slightly to 0.38 ± 0.05 U/mg. The kinetic properties of wild type and its mutants were determined using L-Ile and -KG as substrates. As shown in Table 1 and Supporting Information Figure S4, in the case of L-Ile, the K m values of K152C and T181C were higher than the K m values of wild type, indicating that the affinity for L-Ile decreased slightly. The (k cat /K m ) l-Ile of K152C was only 1.36 s −1 /mM −1 , which was considerably lower than that of wild type (3.18 s −1 /mM −1 ), which could be due to lower catalytic ability of L-Ile. In addition, the (k cat /K m ) l-Ile value of T181C was 1.13-fold that of wild type, which enhanced the catalytic efficiency of L-Ile. In the case of -KG, the K m values of K152C and T181C were lower than that of wild type. Based on the results, the affinity of K152C and T181C for -KG increased. The k cat of T181C was only 0.14 s −1 , which was substantially lower than the k cat of wild type (0.32 s −1 ), which could have been influenced by the catalytic ability of -KG resulting in a further decrease in specific activity. The (k cat /K m ) -KG of T181C was 0.22 s −1 /mM −1 , which was 2.44 times higher  (Figure 3), T220 and R62 are the predicted -KG binding sites. The position of the disulfide bond is near the potential -KG active site; therefore, it may influence specific activity.

Detection of disulfide bond number
To verify the introduction of the disulfide bond into K152C and T181C, we used dithionitrobenzoic acid reagent to determine the contents of free cysteine in the protein. Although a disulfide bond was not introduced in the K152C variant, it was introduced into T181C, whereas no disulfide bond was present in wild type (Table 2).

Structure and molecular dynamics analysis
CD spectroscopic analysis was used to evaluate the effect of the disulfide bond on the IDO secondary structure. The far-UV CD (190 to 260 nm) of the wild type and its mutants revealed a series of similar curves (data not shown) [34,45]. Analysis of the protein CD spectra on the DichroWeb server revealed no significant difference between the secondary structure of wild type and the K152C variant (helix, strand, turns, and unordered) [35]. Rosetta software analysis showed that the ΔG2 for T181C was 71.51 kJ/mol, and the ΔG1 for the wild type IDO was 76.16 kJ/mol. The mutant T181C has a negative ΔΔG (ΔΔG = ΔG2-ΔG1 = −4.65 kJ/mol). The results showed that the introduction of disulfide bond resulted in decreasing the overall folding free energy, thereby improving the thermostability of IDO. Afterward, we simulated the molecular dynamics of wild type and its mutants at 300 K for 30 ns. As shown by the root mean square deviation (RMSD) calculated using the C atom (Figure 2A), all the enzymes exhibited an equilibrium state in the last approximately 13 ns. Generally, an overall decrease in RMSD reflects increased rigidity and in turn increased stability of the enzyme structure [32,43]. The T181C variant had a lower RMSD value than wild type, indicating a more stable structure. With the root mean square fluctuation (RMSF) as an indicator ( Figure 2B), variant T181C had reduced RMSF values at regions 178-187 and 223-231, which increased the stability of loops at C181 and C226, which in turn increased the overall stability of wild type. The RMSF values of wild type in the C61 and K152 region were generally low, indicat-ing the relative stability of the regions. Therefore, mutation did not substantially influence the thermostability at position 152 of IDO. Finally, an RMSF visual analysis of all the atoms in the structure wild type and T181C was performed using VMD ( Figure 3). For T181C, VMD results showed that loops T181 and C226 were bound tightly, which decreased the flexibility of the loop; The RMSF value decreased near regions T181 and C226, which made the overall enzyme structure more compact and increased the rigidity and stability of the structure. Additional evidence of increased structural rigidity was an improvement in the correlation movement [40,41] ( Figure 4). Pairwise cross-correlation coefficients (C ij ) indicate the extent to which the fluctuation of an atom is correlated (or anticorrelated) with another atom, and dynamics cross-correlation maps show the correlation coefficients (C ij ) between all C atom pairs. T181C has less noncorrelation between the amino acid residues compared to wild type, and the correlation between the residues at the 181 and 226 specific sites was increased. The results indicated that the structural stability of the loop region plays a key role in the thermostability of IDO. Therefore, the key role of the C181-C226 disulfide bond in IDO protein thermostability was also supported by the results of molecular dynamics simulation.

4-HIL synthesis by whole cells expressing IDO and T181C
In this study, one-batch whole-cell biotransformation process was applied to B. subtilis 168/pMA5-ido T181C recombinant cells to transform L-Ile to 4-HIL at 30 • C and pH 7.0. The 4-HIL yield increased rapidly within the first 3 h, reaching 134.34 ± 9.68 mM, and the productivity was 6.59 g/(L⋅h). After 9 h, the 4-HIL yield increased gradually. After 21 h, the final yield reached 190.88 ± 5.57 mM, which was a 1.84fold increase compared with the yield from wild type strain (103.84 ± 5.77 mM) ( Figure 5).
Based on the above results, the optimal conversion time was determined to be 9 h for the following repeated batch biotransformation. The B. subtilis 168/pMA5-ido T181C cells were used for five batches. The 4-HIL yield in the fifth batch was still 84.1% that of the first batch ( Figure 5D). After the fifth batch, the final cumulative 4-HIL yield reached 856.91 mM (126.11 g/L), which was a 2.19-fold increase compared with the yield in wild type strain (390.52 mM, 57.47 g/L) ( Figure 5C). The results indicated that the introduction of disulfide bonds enhanced the thermostability and enzyme activity of wild type, hence improved the conversion rate of L-Ile to 4-HIL.
Previously, E. coli 2Δ (ΔsucAB, ΔaceAK, P L -brnQ) was engineered for synthesis of 4-HIL by deleting zwf, edd, and eda genes and expressing B. thuringiensis 2e2 ido gene resulting in approximately 156 mM 4-HIL (22.9 g/L) from 200 mM supplemented L-Ile and self-produced -KG [48]. The ido gene was introduced into C. glutamicum strain SN01 producing L-Ile to synthesize 4-HIL from its own produced L-Ile and -KG [46]. Then, ppc was overexpressed to enhance OAA supply, resulting in production 95.72 mM of 4-HIL after 144 h [15]. Quite recently, dynamically modulated the activity of ODHC by L-Ile-responsive transcription or attenuation strategies, which produced up to 34.21 g/L 4-HIL after 64 h of fermentation [13]. By synergistically promoting the supplies of three substrates (L-Ile, -KG, and O 2 ) and IDO activity, 112-117 mM 4-HIL was obtained after 72 h fermentation in the optimized medium [47]. The main disadvantage of producing 4-HIL by fermentation is that the cycle is long yet the yield is not very high.
Therefore, the production of 4-HIL by biocatalytic has become a trend. By deleting the sucA and aceA genes and expressing a mutant ido gene in E. coli, 151.9 mM 4-HIL (22.4 g/L) was synthesized in 12 h by frozen resting cells when 160 mM L-Ile and 160 mM -KG were added [1]. Expression of a mutant ido gene of Bacillus weihenstephanensis in E. coli, yielded 66.5 mM 4-HIL in 24 h by resting cells when 100 mM L-Ile and 100 mM -KG were added [2].Comparing these previous reports with the final cumulative 4-HIL yield obtained 856.91 mM (126.11 g/L) in this study; we concluded that, it is the highest level of productivity reported based on microbial process. Cell immobilization strategy could be explored at a later stage for industrial production of 4-HIL.

CONCLUDING REMARK S
In the present study, we improved the thermostability and enzyme activity of IDO by the rational design of the disulfide bond. The maximum productivity of whole-cell transformation was increased from 2.53 to 6.59 g/(L⋅h) after mutation, and the final yield increased from 103.84 ± 5.77 to 190.88 ± 5.57 mM, which is the highest productivity ever reported. The whole cell was continuously transformed into five batches of biocatalyst. The cumulative yield of 4-HIL from B. subtilis 168/pMA5-ido T181C reached 856.91 mM (126.11 g/L) at 45 h, which was a 2.19-fold increase compared with the yield of B. subtilis 168/pMA5-ido (390.52 mM, 57.47 g/L). Therefore, the results of the present study could facilitate the industrial production of 4-HIL. This work provides novel insights into the redesigned microreactor, which efficiently produces 4-HIL based on disulfide bond engineering. The approach also paves the way for the potential enhancement of thermostability and enzyme activity of other hydroxylases for the efficient production of other valuable hydroxylated amino acids.